Research On Railway Line Foreign Intrusion Detection Algorithm Based On Machine Vision

With the rapid development of China’s railway industry,China’s railway scientific research level is in a leading position in the world.The operating speed of China’s railways continues to increase,and the scope of operation is gradually covering the whole country,and the railway network is also getting better.However,the subsequent railway safety problems have gradually entered the public’s field of vision.Among the many factors that affect the safety of train operation,the intrusion of foreign things on railway lines is a very important aspect.Therefore,the detection technology of foreign things between tracks has become a very important means to ensure the safety of driving,and it also plays an indispensable role in the rapid development of the railway industry.Aiming at the current research status of foreign things intrusion detection on railway lines at home and abroad,this paper proposes an algorithm to classify foreign things according to their size.The foreign matter that can cover the rail is defined as a large foreign body,and the foreign matter located between the rails is defined as a small foreign body.Different detection schemes have been designed according to the different characteristics of different foreign things.The research content of this article is as follows:(1)In the detection algorithm of large foreign objects,according to the definition of large foreign objects in this article,it can be known that large foreign objects refer to foreign objects that can cover the track.Therefore,the detection of large foreign objects can be transformed into the detection of tracks.In other words,if two complete rails can be detected,and the two rails intersect in a straight line at a distance,there is no large foreign body between the rails,otherwise there is a large foreign body.The edge detection method is used in the algorithm of detecting the rail.Based on the study of traditional Canny operator edge detection and other edge detection methods,this paper has discovered the shortcomings of the Canny operator algorithm in Gaussian filtering,gradient calculation and determination of high and low thresholds,and proposed an improved wavelet Transformation replaces Gaussian filtering to achieve image noise reduction.A new calculation method is used to calculate the gradient,and an improved iterative algorithm is used to adaptively determine the high and low thresholds.After simulation test comparison,the optimized Canny operator algorithm has better detection effect.Then the Hough transform is used to extract the track from the result of the edge detection,and the track is extracted to determine whether there is a large foreign body.(2)In the detection algorithm of small foreign objects,since the definition of small foreign objects is the foreign object located between the rails,that is,the detection target is located between the two rails.The first step of the small foreign body detection algorithm is to extract the track part from the original image,which greatly reduces the workload of image processing;the second step of the small foreign body detection algorithm is to use the optimized ant colony algorithm to detect the extracted track image.The optimization of ant colony algorithm includes two aspects: optimization of path selection probability and optimization of pheromone update method.Then threshold the results of edge detection to obtain the foreign body candidate area;According to the extracted feature parameters,the trained SVM is used for classification processing,so that the small foreign objects between the tracks can be accurately classified and identified.